Self Driving Cars: AI Startup Works To Make Driverless Cars More Like Humans

A start-up called Drive.ai, based in Mountain View, California, is trying to address how an autonomous car would communicate with other drivers and pedestrians. The company is emphasizing what is known in the artificial intelligence field as "human-machine interaction" as a key to confusing road situations.

The autonomous vehicle tech startup found by former graduate students working in Stanford University's Artificial Intelligence Lab, has raised $12 million from an undisclosed venture capital firm and strategic investors. The start-up was forced out of stealth in April when it was awarded a license to test autonomous vehicles in California.

How the tech works

"Everybody talks about this magical world where all the cars on the road are self-driving," says Carol Reiley, cofounder and president of Drive.ai. "It's surprising that the human side has been engineered out of the equation." The company is leaning on a technology called deep learning, a machine-learning technique that is used for a variety of tasks, like understanding human speech and improving the ability to recognize objects in computer vision systems.

"There's the left brain in which a lot of discussion has taken place, what algorithms and what sensors, the logical side," she said. "A lot of the discussion around self-driving cars has no human component, which is really weird because this is the first time a robotic system is going out in the world and interacting with people."

What the cars will be like

According to Fortune, Drive.ai is outfitting its six autonomous test vehicles with sensors, software, and audio equipment to learn how to best communicate with people. An LED sign will be on the roof rack and will be able to send messages and emojis directed at pedestrians and other drivers. The car will also be able to emit sounds.

The car might sound a friendly horn and display a message to tell a pedestrian that it's OK to cross in front of them.The systems will make use of deep learning, a technique that has proven very powerful for teaching machines how to do tasks that would be difficult to program by hand.

The company's first product will be hardware required to retrofit a vehicle so that it can drive itself. It will be offered to companies that operate fleets of vehicles along specific routes, such as delivery or taxi services. Besides sensors and systems for controlling the car, this will include a roof-mounted communications system and a novel in-car interface.

"The sensing and processing difficulties that many of the key technology firms are currently focused on will be solved faster than our ability to devise cohesive human-centered designs," says Bryan Reimer, a research scientist at MIT, through their official news website. "While we will see many deployments of higher order automated vehicles, I believe our roadways will be dominated with lower-level automated systems for decades to come."